58 – Subclassing immutable types

58 – Subclassing immutable types#

How do you subclass immutable types? (For example, how would you subclass floats?)

The dunder method __init__ alone isn’t enough; you need something else…

You need the dunder method __new__, __init__’s big brother. The dunder method __new__ is a class method that is responsible for creating the object, whereas __init__ simply initialises/customises it.

Here is the skeleton for a float subclass:

class FloatSubclass(float):
    def __new__(cls, value, *args, **kwargs):
        print("__new__", value, args, kwargs)
        return super().__new__(cls, value)

    def __init__(self, value, *args, **kwargs):
        print("__init__", value, args, kwargs)
        # Do whatever with the args and kwargs

x = FloatSubclass(4.5)
# __new__ 4.5 ('hello',) {'foo': True}
# __init__ 4.5 ('hello',) {'foo': True}
print(x)  # 4.5 <- looks like a float.